Artificial intelligence is better than students coped with “analogue reasoning”
[ad_1]
The GPT-3 language model did better than students at the University of California, Los Angeles, ranked 12th in the prestigious Shanghai University Rankings, in solving unfamiliar problems by analogy.
Analog reasoning is the solution of unknown problems using analogs. An example of such reasoning is the solution of tests, where several geometric shapes are shown and you need to guess the missing one.
The authors of the work, published in the journal Nature Human Behavior, invited students and artificial intelligence to solve analog reasoning problems. The students scored about 60 percent, and the algorithm scored 80 percent. At the same time, students and GPT-3 made similar mistakes, which suggests a similar decision-making process.
[ad_2]
Source link